Target-based privacy preserving association rule mining.

Association rule mining is an important data mining task applicable across many commercial and scientific domains. There are instances when association analysis must be conducted by a third party over data located at a central point, but updated from several source locations. The source locations may not allow tracking changes. The target location must then take charge of the changed data detection and privatization process. We propose a solution to conduct privacy preserving association rule mining on such data. An evaluation of our approach shows that compared to existing approaches, it renders higher privacy, preserves 90% -100% of the rules and is efficient for 10% database changes.